Smart device stealing and CANDIES

Abstract

Personal devices such as smart phones are increasingly utilized in everyday life. Frequently, activity recognition is performed on these devices to estimate the current user status and trigger automated actions according to the user’s needs. In this article, we focus on improving the self-awareness of such systems in terms of detecting theft: We equip devices with the capabilities to model their own user and to, e.g., alarm the legal owner if an unexpected other person is carrying the device. We gathered 24 h of data in a case study with 14 persons using a Nokia N97 and trained an activity recognition system. Using the data from this study, we investigated several autonomous novelty detection techniques, that ultimately led to the development of CANDIES. The algorithm is able to continuously check if the observed user behavior corresponds to the initial model, triggering an alarm if not. Our evaluations show that the presented methods are highly successful with a theft detection rate of over 85% for the trained set of persons. Comparing the experiments with state of the art techniques support the strong practicality of our approach.